Systematic Review Shows Most Current Readmission Risk Prediction Models have Poor Predictive Ability
Predicting hospital readmission rates is of great interest, both to identify which patients would benefit most from care transition interventions and to standardize readmission risk rates for purposes of hospital comparisons. This systematic review was performed to synthesize the available literature on validated readmission risk prediction models, describe their performance, and assess their suitability for clinical or administrative use. Investigators conducted a literature review of studies on readmission risk prediction models in medical populations that were published from database inception through August 2011. After reviewing nearly 8,000 citations, 30 studies (23 were based on U.S. healthcare data; 4 used VA data) of 26 unique models met the inclusion criteria. Studies that focused on psychiatric, surgical, and pediatric populations were excluded because factors contributing to readmission risk might be very different in these patient groups. For each of the 30 studies, investigators assessed: population characteristics, setting, number of patients in the derivation and validation cohorts, outcomes from use of medical services, readmission rate, range of readmission rates according to predicted risk, and ability of models to discriminate between patients who were subsequently readmitted and those who were not. Fourteen models used retrospective administrative data that could be routinely used for hospital comparisons. The remainder incorporated more detailed clinical and administrative data collected from chart review, primary data collection, or the electronic health record. The most common outcome used was 30-day readmission.
- Most current readmission risk prediction models that were designed for either comparing hospital performance or clinical purposes have poor predictive ability. Although in certain settings such models may prove useful, better approaches are needed to assess hospital performance in discharging patients, as well as to identify patients at greater risk of preventable readmission.
- Most models incorporated variables for medical comorbidity and use of prior medical services, but few examined variables associated with overall health and function, illness severity, or social determinants of health. The variable performance of predictive models in different populations suggests that the best choice of a model may depend on the setting and population in which it is being used.
- Even though the overall predictive ability of the clinical models was poor, investigators found that high- and low-risk scores were associated with a clinically meaningful gradient of readmission rates. Thus, even limited ability to identify a proportion of patients at highest risk for readmission could increase the cost-effectiveness of hospital interventions aimed at improving the discharge process and post-hospital follow-up.
Additional research is needed to assess the true preventability of readmissions in U.S. health systems.
The poor performance of models raises concerns about the ability to broadly compare risk-standardized readmission rates across hospitals in a fair manner.
- Studies outside of the U.S. were included; applicability of these models may be limited.
- Few studies directly compared models within the same population, and summary statistics should not be used to directly compare models across different populations.
This study was partly funded by HSR&D (ESP 05-225). Dr. Kansagara is part of VA’s Evidence-Based Practice Center in Portland.
Kansagara D, Englander H, Salanitro A, et al. Risk Prediction Models for Hospital Readmission: A Systematic Review. JAMA October 19, 2011;306(15):1688-98.